=== chihchun_afk is now known as chihchun [06:16] Good morning [07:14] jibel: !! [07:14] $ ./run-from-checkout -d ~/ubuntu/tmp/ubuntu-calculator-app/ ~/ubuntu/tmp/com.ubuntu.calculator_1.3.283_all.click --- ssh -s ssh-setup/adb [07:14] adt-run [09:14:23]: test autopilot: - - - - - - - - - - results - - - - - - - - - - [07:14] autopilot PASS [07:18] pitti, \o/ [07:19] jibel: still with one remaining hack, but I committed the "advertise suggested-normal-user=" capability fix which was missing [07:22] jibel: did you ever happen to figure out how to run powerd-cli in the background? [07:22] --setup-commands "(powerd-cli display on bright &) &" [07:22] even with this double fork it hangs eternally [07:45] pitti, usually I do "powerd-cli display on & " it hangs over adb but not with ssh [07:49] jibel: that still hangs for me [08:27] jibel: ugly, but working: http://paste.ubuntu.com/7725346/ [08:31] pitti, ugly really :) [08:32] pitti, you'd probably want to force bash for disown. it is not a builtin of dash and I think adb shell calls sh -c [08:34] jibel: hm, it seems to call bash here [08:34] jibel: but adding another "bash -c" probably can't hurt [08:43] jibel: now I just need to disable the screen lock [08:55] pitti, BTW I'm sure you noticed bug 1335176 [08:55] bug 1335176 in autopkgtest (Ubuntu) "Add man page for ssh runner" [Undecided,New] https://launchpad.net/bugs/1335176 [08:55] just confirming [08:55] jibel: yes, I did; next thing after I add the aa-clickhook bits [08:56] jibel: btw, might be easier to put your own /users git branch on alioth, and/or asking for membership in the autopkgtest alioth project [08:56] s/branch/repository/ [09:16] pitti, for unlocking the device, there is a helper in unity8-autopilot [09:16] pitti, http://bazaar.launchpad.net/~unity-team/unity8/trunk/view/head:/tools/unlock-device [09:17] jibel: thanks; vila just pointed that out in #touch [09:17] ok === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === davmor2_ is now known as davmor2 [10:03] Morning all === rfinley is now known as average_guy [11:05] jibel: looking at (and simplifying) your nova script now; that doesn't have the floating-ip-create/floating-ip-associate stuff, isn't it necessary in general or just not for the canonistack? [11:20] pitti, I didn't try with hpcloud, but the private ip should be enoguht [11:20] enought [11:20] -t [11:21] | evan.dandrea@canonical.com-network network | 10.0.0.6 [11:21] jibel: ^ that's what I get without a floating IP [11:21] pitti, right and can you ssh to 10.0.0.6? [11:22] jibel: no, as I said I need to add a public floating one [11:22] at least I haven't figured out how to use the 10.0.0.x one, supposedly you need some kind of VPN to the HP cloud? [11:22] ev, vila: ^ did you happen to figure out how to use these "private" IPs from HP cloud? [11:23] i. e. without floating-ip-create/floating-ip-associate ? [11:27] pitti, ah, you're right, we probably cannot for hpcloud. For canonistack I'm forwarding via chinstrap [11:37] jibel: detecting this automatically is a nuisance, so I suppose I'll just add an option to create/associate a floating IP [11:46] pitti: you need a floating IP, juju may handle it for you otherwise you're on your own [11:46] vila: *nod* === fader_` is now known as fader [12:44] jibel: so with canonistack you can immediately ssh in after nova boot? with HP cloud it still takes a minute or so until ssh actually works, so I need to add an ssh wait loop [12:46] jibel: oh, that's already in adt-virt-ssh now, nevermind [12:46] HTTPSConnectionPool(host='region-a.geo-1.compute.hpcloudsvc.com', port=443): Max retries exceeded with url: /v2/11490006884368/servers (Caused by : [Errno 104] Connection reset by peer) [12:46] ERROR: Unable to delete any of the specified servers. [12:46] and I'm getting fun like that [12:47] pitti, jibel: you're running into problems that has been solved in uci-engine testbed.by (floating IP, slow boot, transient hp errors, etc) [12:48] vila: ah, ncie [12:48] vila: so I suppose for CI we should use an ssh setup script which uses uci [12:48] pitti: exactly [12:49] vila: you mentioned a script/package that sets these up, uci-vm or something? [12:50] pitti: uci-vms yes, but the nova part is still uci-engine specific for now (and we're still trying to find the right cloud config with webops... can't use nova in prodstack :-/) [12:51] vila: ack; the ssh setup scripts will be highly environment specific anyway, I mostly want to include some of them (adb, nova) as examples; although the adb one works quite nicely now [12:51] vila: so as soon as uci-vms is packaged somewhere, we can also include an ssh setup script which uses that [12:52] vila: while developers will want to use the adb setup script to reproduce errors locally, I suppose they'd rather use adt-virt-qemu for reproducing a package test failure than ssh with nova [12:52] few people actually have access to the HP (or other) cloud, and even if you do, it's a magnitude slower [12:53] vila: so ATM the nova script is mostly just good enough to prove that adt-virt-ssh can deal with this kind of testbed [12:54] pitti: indeed, I don't want to stop you from exploring, just mentioning that hpcloud can be a can of worms (well, so can canonistack ;-) [12:54] pitti: yup [12:54] heh, amen [12:54] vila: so to be on the same page, you are mostly looking for the ssh runner itself, but want to write your own setup script using uci-vms? [12:57] vila: so that I should make sure that the ssh runner provides enough functionality, but the shipped nova setup script doesn't need to be production-stable? [12:57] pitti: exactly [12:58] vila: très bien; then I won't waste too much time on it, let's rather package uci-vms at some point and then use that [12:58] pitti: IP, login, ssh key [12:58] and treat it as a PoC [12:59] pitti: it's packaged but only available in https://launchpad.net/~canonical-ci-engineering/+archive/ci-airline-phase-0 for now [12:59] pitti: feedback welcome on the packaging in any case ;) [13:00] jibel: hmm, current nova script doesn't export "identity", how did that work for you? [13:00] jibel: also, I'm not entirely sure how to find the private key now, when just giving the keypair name; keypair-show only shows the pub key, but not its location [13:01] pitti, as I said I just tried with canonistack and I had my account already setup. So probably better rename nova -> canonistack and just as a proof of concept than anything else :) [13:02] jibel: right, but even with canonistack the ssh runner shoudl just fail without an "identity="? [13:02] pitti, I can have a look later this week with another cloud [13:02] jibel: no worries, I'll add it [13:03] jibel: or perhaps rather just default to ~/.ssh/id_rsa.pub in virt-ssh, seems easier than doing it for every script [13:04] pitti, in my ssh/config I have Host XX.XX.XX.XX [13:04] User ubuntu [13:04] jibel: aah [13:04] IdnetityFile [13:04] and the proxy command for chinstrap [13:05] jibel: hm, true, if the script doesn't give identity=, ssh should default to the usual key by itself, so nevermind === plars_ is now known as plars [13:12] jibel: nevermind, it was a bug in wait_for_ssh(), fixed now [13:12] jibel: now I get much further, I just run into unexpected stderr "sudo: unable to resolve host adt-nova-6ogg8l" [13:13] jibel: almost there :) [13:16] pitti, sorry for the bugs [13:16] jibel: no worries [13:17] jibel: just pondering how to set the host name properly, in the setup script or in virt-ssh [13:19] pitti, IMO in the setup script because usually it's already setup correctly. For example, on a phablet you won't change it, on an existing host it's already done, on canonistack cloud-init takes care of it, on lxc it's done in the template, ... [13:20] it seems to be really target specific [13:20] jibel: yeah, I agree [13:20] jibel: but I won't want to replicate the ssh wait loop, so I'm now reading http://docs.openstack.org/user-guide/content/user-data.html how to do this kind of setup at "nova boot" time instead of through ssh [13:21] --meta Record arbitrary key/value metadata to /meta.js on the [13:21] new server. Can be specified multiple times. [13:21] that looks promising, if cloud-init can use that to fix /etc/hosts [13:22] (but this is really just a bug in cloud-init or the HP version of it, *grump*) [13:22] vila: ^ does uci-vms magically happen to fix that, too by chance? [13:26] pitti: sorry, which one ? host name ? yeah, through cloud-init meta data, fuzzy memory there, I can't remember when I ran into issues there, I don't think I use it for nova though [13:26] vila: creating an instance foo-123 doesn't seem to add "foo-123" (the hostname) to /etc/hosts by default, so that e. g. sudo complains (apache and postfix will as well) [13:26] I suppose I want to set manage_etc_hosts [13:27] meh, this is really stuff that ought to "just work" [13:28] pitti: hmm, I remember seeing that on juju instances but not on testbeds (and indeed I do use manage_etc_hosts) [13:28] pitti: in uci-vms that is [13:31] meh, sometimes it feels like all that stuff makes it specifically hard to be used [13:31] so there is a --meta to set an individual key, but not a corresponding one for setting a cloud-config option !? [13:31] just a whole --user-data thing, but thanks no, I don't want to manage the *entire* configuration myself [14:02] jibel: I just pushed the revised nova script, which now also works for HP; would you mind testing adt-virt-ssh HEAD again against canonistack? [14:05] jibel: finished the libpng test, in a whopping 7:06 minutes (compare to schroot runner's 4 seconds..) [14:07] pitti, OTOH you can boot thousands in parallel :) [14:07] *of VMs [14:07] I'll test it in canonistack [14:07] yeah, still much room for optimization :) [14:08] jibel: ssh connnection sharing might help quite a bit with adb though, I'll still look into that [14:08] jibel: but many thanks for all your ground work here, nice to see all the puzzle pieces now [14:08] pitti, I tried quickly but the runner exited with an error 255 from ssh. If I try interactively it is fine. [14:15] jibel: do you have a -d (with both) output? [14:16] pitti, I don't. [14:16] I'll try again. === bdmurray_ is now known as bdmurray [15:55] robotfuel: how can I run the tests you are working on? [15:59] elopio: the test is here https://code.launchpad.net/~chris.gagnon/+junk/lrt [16:00] robotfuel: branching... === chihchun is now known as chihchun_afk [16:00] robotfuel: why is it on your junk? maybe inside launchpad.net/ubuntu_autopilot_tests would be a good place for it. [16:02] elopio: ok [16:03] robotfuel: ok, what we will need in order to have it running for every image is a deb package.cd .. [16:04] robotfuel: do you know about packaging? [16:04] elopio: I use run-lrt.sh to start the test because launchpadlib is py2 and the other code is in py3.. so I can upload the bugs [16:04] elopio: yes, I can package it [16:05] robotfuel: and then we will have to add it to the archive: https://wiki.ubuntu.com/UbuntuDevelopment/NewPackages#Going_through_MOTU [16:05] I'm doing the same for ubuntu_experience_tests, so it would be nice to do it together :D [16:05] sure [16:07] elopio: my plan for today is get the graph working for time to failure. then I will do the packaging and add more types of lrt tests. [16:07] ok, thanks. robotfuel: let me know if you need a hand or a review. [16:07] * elopio goes to get breakfast. === om26er is now known as om26er|away === om26er|away is now known as om26er === chihchun_afk is now known as chihchun === chihchun is now known as chihchun_afk [17:47] bdmurray: is there a way to avoid the crichton bot to subscribe ~ubuntu-sponsor and add the tag patch on bugs where someone upload a debdiff? maybe a particular tag? [17:53] mapreri: 'bot-stop-nagging' should prevent that from happening [17:54] bdmurray: ok, thanks [18:27] I'll upgrade my quassel server to trusty because it's unbearable. I might be away for some time. === chihchunl is now known as chihchun === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === salem_ is now known as _salem